Efficient Algorithms Of Resource Allocation In Virtual Data Centers

Virtual Data Center(VDC)technology can efficiently improve the resource utilization through virtualizing physical resources in a data center.VDC resource embedding is very important for realizing the VDC technology.Although there have been many studies on VDC(resource)embedding,they mainly focus on unicast services,but do not consider the mixture of unicast and multicast services.Moreover,these studies are also based on the assumption of a fixed capacity of each VDC,but do not consider the time-varying feature of the capacity required.In a real data center,there exists both unicast and multicast services.Also,the capacity of each VDC may change according to different service requirements.In this thesis,we first propose the concept of mixed VDC embedding to support the mixture of unicast and multicast services.Moreover,in order to improve the embedding efficiency,we incorporate multicast capabilities in a data center to support multicast services.We develop a Mixed Integer Linear Programming(MILP)model and an efficient heuristic algorithm for the mixed VDC embedding problem.Numerical results show that the proposed mixed VDC embedding scheme can significantly reduce system cost,energy consumption,and network resources used under the scenario of static VDC requests,and significantly reduce the VDC request blocking ratio under the scenario of dynamic VDC requests.In addition,considering the time-varying feature of VDC resources required,we first predict the resource requirement of each VDC at different moments using the Artificial Intelligence(AI)technology.We also propose a new VDC consolidation scheme,which embeds VDCs whose peaks and valleys of resource demands stagger with each other,onto common servers and physical links.An Integer Linear Programming(ILP)model and a correlation-based heuristic algorithm are also developed to verify the efficiency of the proposed approach.